康普吞相機(Compton camera)可以偵測重離子/質子誘發產生的瞬發加馬光子，並能於重離子/質子治療過程提供在線式非侵入性的三維質子射程成像驗證。瞬發加馬光子僅於重離子/質子射束照射時產生，其能量分布廣且可高達數個MeV，因此對重離子/質子治療應用領域，康普吞相機的偵檢效率將是一大挑戰。
本研究設計並研製側向式康普吞散射層偵檢器，由1.8 × 1.8 × 100 mm 的矽酸釔鑥閃爍晶體(Lutetium-yttrium oxyorthosilicate, 簡稱LYSO)組成並透過雙端的矽光電倍增器(silicon photomultiplier)讀出。高原子序的LYSO閃爍晶體可增加康普吞散射偵檢效率，而其快速時間響應可降低非真實事件的偵測。此偵檢器使用非常長且細的晶體，可確保提供康普吞相機足夠的視野範圍(Field of View, 簡稱FOV)與射程深度解析度。但細長型的晶體會造成長軸向的空間解析度與能量解析度下降。為改善此問題，我們提出一種新偵檢器構型(configuration)，利用LS-NN—光分享(Light Sharing, 簡稱LS)設計搭配神經網路(Neural Network, 簡稱NN) 的位置解碼方法，提升康普吞相機之散射層偵檢器性能。
首先以蒙地卡羅模擬比較AS (Arrow Shape)、ITS (Isosceles –Triangular Shape)與OTS (Opposite-triangular Shape)三種不同的光分享設計。模擬結果顯示AS光分享構型具有最佳的性能，因此採用AS構型進一步實作偵檢器並實測評估其長軸向的空間解析度、能量解析度及晶體響應圖品質。另外，亦實作不具光分布的傳統FC (Fully Covered)構型為對照組以進行性能比較。量測結果顯示AS光分享構
型搭配LS-NN法可使100 mm長LYSO偵檢器的長軸向空間解析度，從21.6 ± 3.4 mm減小至9.2 ± 3.1 mm，提升了57.4%。值得一提的是這個方法不僅具空間解析度優勢，也能使偵檢器能量解析度提升15.3%。然而光分享構型會稍微劣化偵檢器的晶體響應圖品質，但該影響相當輕微可忽略。
本研究提出創新的LS-NN加馬偵檢探頭設計，透過實作與度量實測，證明本設計可以同時提升偵檢探頭的空間與能量解析度。這個設計方便易實現且具性能優勢，在重離子/質子治療用在線式康普吞成像系統的側向式散射層探頭開發應用上具有很大潛力。

Heavy ion/proton-induced prompt gammas can be detected by a Compton camera, which provides in vivo on-line imaging for heavy ion/proton range verification. Since the prompt gammas emitted during the beam-on time are with energies up to several MeV, an on-line prompt gamma Compton camera with high detection efficiency is challenging to implement.
In this study, a side-on scatterer of Compton camera was designed and developed. It was composed of 1.8 × 1.8 × 100 mm lutetium-yttrium oxyorthosilicate (LYSO) crystals with dual-ended readout by silicon photomultipliers (SiPMs). The side-on scatterer using very long and thin crystals is to ensure the Compton camera’s field of view and the spatial resolution in the heavy ion/proton-depth direction. However, the very long and thin crystals will cause degradation of lateral spatial resolution and energy resolution of the side-on sctterer. Therefore, we proposed a novel light sharing-neural network (LS-NN) method that combines with light sharing configuration and neural network position decoding algorithm to improve the spatial resolution and energy resolution for the side-on scatterer.
Three light sharing configurations, the AS (Arrow shape), ITS (Isosceles-triangular Shape), and OTS (Opposite-triangular Shape) were modeled using Monte Carlo simulation. Simulation results show that the AS configuration provided the best performance than others. Therefore,
the AS configuration was constructed for measurement to evaluate the long-axial spatial resolution, energy resolution, and the quality of the flood histogram of the side-scatterer. A conventional FC (Fully Covered) configuration without light sharing was built for comparison. The measurement results show that the proposed LS-NN method can give 57.4% and 15.3% improvements in the long-axial spatial and energy resolution for the side-on scatterer, respectively. As for the quality of flood histogram of the scatterer, only a modest degradation was observed for the light sharing configuration.
A new gamma imaging detector employing very long and thin crystals was designed and implemented. With the proposed learning-based, LS-NN algorithm, we demonstrated that both the position-of-interaction resolving capability and energy performance of thedetector were significantly improved. It is concluded that the proposed LS-NN with AS configuration is a potential candidate as a Compton scatterer, which is promising for developing a high-performance prompt gamma imaging system for range verification of proton and ion beams.

摘 要 i
Abstract iii
致 謝 v
目 錄 vi
圖目錄 viii
表目錄 xii
第1章 緒論 - 1 -
1.1 研究背景 - 1 -
1.2 研究目的 - 1 -
1.3 論文架構 - 3 -
第2章 文獻回顧 - 4 -
2.1 粒子治療與粒子射程驗證問題 - 4 -
2.2 間接偵測質子射程的方法 - 4 -
2.3 正子互毀加馬光子偵測 - 6 -
2.4 瞬發加馬光子偵測 - 7 -
2.4.1 非影像系統 - 8 -
2.4.2 影像系統 - 11 -
2.5 閃爍晶體康普吞相機 - 14 -
2.6 加馬光子作用閃爍晶體的位置資訊 - 16 -
2.6.1 單晶(monolithic crystal) - 16 -
2.6.2 像素化的晶體(pixelated crystal) - 16 -
2.6.3 多層晶體(multi-layer crystal) - 18 -
2.7 深度學習(deep learning) - 19 -
2.7.1 監督與非監督式學習 - 20 -
2.7.2 反向傳遞演算法 - 21 -
2.7.3 優化器 - 22 -
2.7.4 過擬合現象(overfitting) - 22 -
第3章 材料與方法 - 24 -
3.1 閃爍體偵檢器 - 24 -
3.2 矽光電倍增器(Silicon Photomultiplier, SiPM) - 26 -
3.3 側向式康普吞相機設計 - 28 -
3.3.1 非真實事件 - 28 -
3.3.2 視野範圍(Field of View, FOV) - 30 -
3.4 偵檢器的加馬作用位置演算法 - 30 -
3.4.1 加馬作用位置估算 - 30 -
3.4.2 光分享神經網路LS-NN - 32 -
3.5 光分享晶體構型 - 33 -
3.6 蒙地卡羅模擬設置 - 34 -
3.7 度量實驗設置 - 38 -
3.7.1 移動平台設置以及射源配置 - 39 -
3.7.2 數據擷取系統 - 41 -
3.8 模擬與度量數據用的神經網路 - 43 -
3.9 數據分析評估方法 - 44 -
3.9.1 偵檢器加馬作用的三維位置估算 - 44 -
3.9.2 POIx解析度 - 44 -
3.9.3 能量解析度(energy resolution) - 45 -
3.9.4 位置誤差(positioning error) - 46 -
3.9.5 峰谷比值(Peak-to-valley ratio, PVR) - 46 -
第4章 結果與討論 - 48 -
4.1 模擬數據結果與分析 - 48 -
4.1.1 AS光分享構型 - 48 -
4.1.2 ITS光分享構型 - 54 -
4.1.3 OTS光分享構型 - 59 -
4.1.4 模擬數據比較 - 64 -
4.2 度量實驗結果與分析 - 66 -
4.2.1 AS光分享構型 - 66 -
4.2.2 FC構型(無光分享) - 70 -
4.2.3 POIx位置估算比較 - 73 -
4.2.4 能量解析度比較 - 73 -
4.2.5 PVR比較 - 75 -
4.3 討論 - 77 -
4.3.1 新型散射偵檢器性能總論 - 77 -
4.3.2 過度擬合問題 - 77 -
4.3.3 未來研究建議 - 80 -
第5章 結論 - 82 -
參考文獻 - 83 -

[1] World Health Organization. Cancer, Available:https://www.who.int/news-room/fact-sheets/detail/cancer.
[2] 衛生福利部統計處. 106 年國人死因統計結果, Available:https://www.mohw.gov.tw/cp-3795-41794-1.html.
[3] B. Stewart and C. P. Wild, "World cancer report 2014," 2014.
[4] F. Stichelbaut and Y. Jongen, "Verification of the proton beams position in the patient by the detection of prompt gamma-rays emission," 39th Meeting of the Particle Therapy Co-Operative Group, 2003.
[5] J. Krimmer, D. Dauvergne, J. Létang, and É. Testa, "Prompt-gamma monitoring in hadrontherapy: A review," Nucl Instrum Methods Phys Res A, vol. 878, pp. 58-73, 2018.
[6] A.-C. Knopf and A. Lomax, "In vivo proton range verification: a review," Phys Med Biol, vol. 58, no. 15, pp. 131-160, 2013.
[7] J. C. Polf and K. Parodi, "Imaging particle beams for cancer treatment," Phys Today, vol. 68, no. 10, p. 28, 2015.
[8] P. Crespo, G. Shakirin, F. Fiedler, W. Enghardt, and A. Wagner, "Direct time-of-flight for quantitative, real-time in-beam PET: a concept and feasibility study," Phys Med Biol, vol. 52, no. 23, pp. 6795-6811, 2007.
[9] T. Yamaya et al., "A proposal of an open PET geometry," Phys Med Biol, vol. 53, no. 3, pp. 757-773, 2008.
[10] T. Nishio, A. Miyatake, T. Ogino, K. Nakagawa, N. Saijo, and H. Esumi, "The development and clinical use of a beam ON-LINE PET system mounted on a rotating gantry port in proton therapy," Int J Radiat Oncol, Biol, Phys, vol. 76, pp. 277-286, 2010.
[11] M. Moteabbed, S. España, and H. Paganetti, "Monte Carlo patient study on the comparison of prompt gamma and PET imaging for range verification in proton therapy," Phys Med Biol, vol. 56, no. 4, pp. 1063-1082, 2011.
[12] S. España and H. Paganetti, "The impact of uncertainties in the CT conversion algorithm when predicting proton beam ranges in patients from dose and PET-activity distributions," Phys Med Biol, vol. 55, no. 24, pp. 7557-7571, 2010.
[13] A.-C. Knopf et al., "Accuracy of proton beam range verification using post-treatment positron emission tomography/computed tomography as function of treatment site," Int J Radiat Oncol, Biol, Phys, vol. 79, no. 1, pp. 297-304, 2011.
[14] W. D. Newhauser and R. Zhang, "The physics of proton therapy," Phys Med Biol, vol. 60, p. R155, 2015.
[15] C. H. Min, H. R. Lee, C. H. Kim, and S. B. Lee, "Development of array‐type prompt gamma measurement system for in vivo range verification in proton therapy," Med Phys, vol. 39, no. 4, pp. 2100-2107, 2012.
[16] J. H. Park et al., "Comparison of knife-edge and multi-slit camera for proton beam range verification by Monte Carlo simulation," Nucl Eng Technol, vol. 51, no. 2, pp. 533-538, 2019.
[17] V. Bom, L. Joulaeizadeh, and F. Beekman, "Real-time prompt gamma monitoring in spot-scanning proton therapy using imaging through a knife-edge-shaped slit," Phys Med Biol, vol. 57, no. 2, pp. 297-308 , 2011.
[18] Y. Xie et al., "Prompt gamma imaging for in vivo range verification of pencil beam scanning proton therapy," Int J Radiat Oncol, Biol, Phys, vol. 99, no. 1, pp. 210-218, 2017.
[19] F. Hueso-González et al., "First test of the prompt gamma ray timing method with heterogeneous targets at a clinical proton therapy facility," Phys Med Biol, vol. 60, no. 16, pp. 6247-6272, 2015.
[20] C. Golnik et al., "Range assessment in particle therapy based on prompt γ-ray timing measurements," Phys Med Biol, vol. 59, no. 18, pp. 5399-5422 , 2014.
[21] J. Krimmer et al., "A cost-effective monitoring technique in particle therapy via uncollimated prompt gamma peak integration," Appl Phys Lett, vol. 110, no. 15 , 2017.
[22] J. M. Verburg and J. Seco, "Proton range verification through prompt gamma-ray spectroscopy," Phys Med Biol, vol. 59, no. 23, pp. 7089-7106 , 2014.
[23] M. Singh, "An electronically collimated gamma camera for single photon emission computed tomography. Part I: Theoretical considerations and design criteria," Med Phys, vol. 10, no. 4, pp. 421-427, 1983.
[24] V. Venugopal and P. S. Bhagdikar, "de Broglie Wavelength and Frequency of the Scattered Electrons in Compton Effect," arXiv preprint arXiv:1202.4572, 2012.
[25] M.-H. Richard et al., "Design study of a Compton camera for prompt γ imaging during ion beam therapy," IEEE Nucl Sci Symp Conf Record, pp. 4172-4175, 2009.
[26] S. Peterson, D. Robertson, J. J. P. i. M. Polf, and Biology, "Optimizing a three-stage Compton camera for measuring prompt gamma rays emitted during proton radiotherapy," Phys Med Biol, vol. 55, no. 22, pp. 6841-6856 , 2010.
[27] A. Kishimoto, J. Kataoka, T. Nishiyama, T. Taya, and S. Kabuki, "Demonstration of three-dimensional imaging based on handheld Compton camera," J Instrum, vol. 10, no. 11, 2015.
[28] Y. Du, Z. He, G. Knoll, D. Wehe, and W. Li, "Evaluation of a Compton scattering camera using 3-D position sensitive CdZnTe detectors," Nucl Instrum Methods Phys Res A, vol. 457, no. 1-2, pp. 203-211, 2001.
[29] M. Ito, S. J. Hong, and J. S. Lee, "Positron emission tomography (PET) detectors with depth-of-interaction (DOI) capability," Biomed Eng Lett, vol. 1, no. 2, pp. 70-81 , 2011.
[30] X. Li, C. Lockhart, T. K. Lewellen, and R. S. Miyaoka, "A high resolution, monolithic crystal, PET/MRI detector with DOI positioning capability," 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, pp. 2287-2290, 2008.
[31] W. Yonggang, C. Xinyi, and L. Deng, "Depth of interaction estimation using artificial neural network for continuous crystal PET detector," IEEE Nucl Sci Symp Med Imaging Conf Rec, pp. 1260-1263, 2012.
[32] R. Marcinkowski, P. Mollet, R. Van Holen, and S. Vandenberghe, "Sub-millimetre DOI detector based on monolithic LYSO and digital SiPM for a dedicated small-animal PET system," Phys Med Biol, vol. 61, no. 5, pp. 2196-2212, 2016.
[33] A. González-Montoro et al., "Performance study of a large monolithic LYSO PET detector with accurate photon DOI using retroreflector layers," IEEE Trans Radiat Plasma Med Sci, vol. 1, no. 3, pp. 229-237, 2017.
[34] Y.-C. Tai and A. L. Goertzen, "Optical Simulation of Dual-Ended Readout of Axially-Oriented 100 mm Long LYSO Crystals for Use in a Compact PET System," IEEE Tran Nucl Sci, vol. 61, no. 1, pp. 3-13, 2013.
[35] A. Mohammadi et al., "Development of a dual-ended readout detector with segmented crystal bars made using a subsurface laser engraving technique," Phys Med Biol, vol. 63, no. 2, 2018.
[36] Z. Kuang et al., "Development of depth encoding small animal PET detectors using dual‐ended readout of pixelated scintillator arrays with SiPMs," Med phys, vol. 45, no. 2, pp. 613-621, 2018.
[37] A. R. Selfridge, S. R. Cherry, and M. S. Judenhofer, "Optimization of a depth of interaction encoding PET block detector for a PET/MRI insert," Phys Med Biol, vol. 63, no. 23, 2018.
[38] G. Stringhini, M. Pizzichemi, A. Ghezzi, A. Stojkovic, M. Paganoni, and E. Auffray, "Development and evaluation of a practical method to measure the Depth of Interaction function for a single side readout PET detector," J Instrum, vol. 11, no. 11, 2016.
[39] M. S. Lee and J. S. Lee, "Depth-of-interaction measurement in a single-layer crystal array with a single-ended readout using digital silicon photomultiplier," Phys Med Biol, vol. 60, no. 16, pp. 6495-6514, 2015.
[40] J.-W. Son, M. S. Lee, and J. S. Lee, "A depth-of-interaction PET detector using a stair-shaped reflector arrangement and a single-ended scintillation light readout," Phys Med Biol, vol. 62, pp. 465-83, 2017.
[41] P. Fan, T. Ma, Q. Wei, R. Yao, Y. Liu, and S. Wang, "Choice of crystal surface finishing for a dual-ended readout depth-of-interaction (DOI) detector," Phys Med Biol, vol. 61, no. 3, pp. 1041-56, Feb 7 2016.
[42] Z. H. Kuang et al., "A depth-encoding PET detector that uses light sharing and single-ended readout with silicon photomultipliers," Phys Med Biol, vol. 63, no. 4, Feb 2018.
[43] J. P. Schmall, S. Surti, and J. S. Karp, "Characterization of stacked-crystal PET detector designs for measurement of both TOF and DOI," Phys Med Biol, vol. 60, no. 9, pp. 3549-3565, 2015.
[44] H. Behnamian and M. Shafiee, "Characterization of three layer depth of interaction PET detectors for small animal imaging," Biomed Phys Eng Express, vol. 4, no. 5, 2018.
[45] H. Liu, T. Omura, M. Watanabe, and T. Yamashita, "Development of a depth of interaction detector for γ-rays," Nucl Instrum Methods Phys Res A, vol. 459, no. 1-2, pp. 182-190, 2001.
[46] N. Inadama, Y. Hirano, F. Nishikido, H. Murayama, and T. Yamaya, "Development of a DOI PET Detector Having the Structure of the X'tal Cube Extended in One Direction," (in English), IEEE Tran Nucl Sci, vol. 63, no. 5, pp. 2509-2516, Oct 2016.
[47] A. Koide et al., "Precision imaging of 4.4 MeV gamma rays using a 3-D position sensitive Compton camera," Sci rep, vol. 8, no. 1, 2018.
[48] D. E. Rumelhart, G. E. Hinton, and R. J. Williams, "Learning internal representations by error propagation," California Univ San Diego La Jolla Inst for Cognitive Science, 1985.
[49] G. E. Hinton and R. R. Salakhutdinov, "Reducing the dimensionality of data with neural networks," Science, vol. 313, no. 5786, pp. 504-507, 2006.
[50] Y. LeCun, Y. Bengio, and G. Hinton, "Deep learning," Nature, vol. 521, no. 7553, pp. 436-444, 2015.
[51] S. Ruder, "An overview of gradient descent optimization algorithms," arXiv preprint arXiv:1609.04747, 2016.
[52] Visualizing Optimization Algos, Available:https://imgur.com/a/Hqolp.
[53] Wikipedia. Overfitting. Available:https://en.wikipedia.org/wiki/Overfitting.
[54] A. Y. Ng, "Feature selection, L 1 vs. L 2 regularization, and rotational invariance," Proceedings of the twenty-first international conference on Machine learning, 2004
[55] N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, "Dropout: a simple way to prevent neural networks from overfitting," J Mach Learn Res, vol. 15, no. 1, pp. 1929-1958, 2014.
[56] E. B. Podgoršak, Radiation physics for medical physicists. Springer, 2006.
[57] Advatech, Radiation Detectors, Available:https://www.advatech-uk.co.uk/index.html.
[58] Hamamatsu. Layout of Hamamatsu silicon photomultiplier, Available:http://www.hamamatsu.com/sp/hc/osh/hub_04e_12_tn1_figure01.jpg.
[59] P. D. Olcott, J. A. Talcott, C. S. Levin, F. Habte, and A. M. Foudray, "Compact readout electronics for position sensitive photomultiplier tubes," IEEE Tran Nucl Sci, vol. 52, no. 1, pp. 21-27, 2005.
[60] C. S. Levin, "New imaging technologies to enhance the molecular sensitivity of positron emission tomography," Proc IEEE, vol. 96, no. 3, pp. 439-467, 2008.
[61] E. Roncali, M. Stockhoff, and S. R. Cherry, "An integrated model of scintillator-reflector properties for advanced simulations of optical transport," Phys Med Biol, vol. 62, no. 12, pp. 4811-4830, 2017.